Use Turf.js for 200-voter Travis turf boundaries and routes

## Problem
The current turf generation uses deterministic chunk geometry and rectangular boundaries. The original plan called for using Turf.js to break Travis County into many 200-voter turfs.

## Goal
Use Turf.js geospatial helpers to create better turf boundaries and walk routes from household point collections.

## Suggested Approach
- Add a Node-side Turf.js materializer or generated artifact step.
- Build point features for geocoded household targets.
- Partition into roughly 200-voter turf groups.
- Use Turf.js helpers such as bbox/center/convex/voronoi or clustering utilities as appropriate.
- Store generated boundaries and route waypoints in SpacetimeDB.

## Acceptance Criteria
- Turf boundaries are generated from household coordinates rather than simple fallback rectangles.
- Each turf is close to 200 registered voters except natural remainder/edge cases.
- Routes are ordered enough for simulated walkers to move naturally through the turf.
- Generation is deterministic and testable.

Start by locating the current turf-generation implementation, the Node-side materializer or generated-artifact step, and the SpacetimeDB storage path mentioned in the issue. Define how household point features are partitioned and routes are ordered, then verify deterministic generation, near-200-voter groups, coordinate-based boundaries, and stored route waypoints against the acceptance criteria.

Written by the indexing model from the issue text.